I spend a lot of time watching how people use AI. And I keep seeing the same pattern:
Someone needs to write a strategy doc. They fire up ChatGPT or Claude. They type “Write a go-to-market strategy for…” Hit enter. Copy. Paste. Done.
They think they’re using AI effectively.
They’re not. They’re using AI like a faster Google Docs, and it’s much better than that.
Here’s what almost nobody is doing: Flipping the script. Having AI interrogate them before generating anything.
That subtle shift, from commanding AI to being challenged by it, separates people who are “busy with AI” from those who are genuinely better because of it.
We Don’t Need More Answers. We Need Better Questions.
The dominant workflow is simple: You have a task. You prompt an LLM. It gives you an answer.
PowerPoint deck. Draft email. Market analysis. Policy summary.
The problem? This only scales what you already think you need. It doesn’t challenge your assumptions, expose your blind spots, or force you to make your thinking explicit.
It makes you faster without making you smarter. Follow me here…
But flip the interaction, and everything changes.
Instead of “write this for me,” you say:
“Before you generate anything, ask me the questions a tough stakeholder would ask. Challenge my assumptions. Make me defend this.” You will be amazed at the better results you get!
Now the LLM isn’t just executing tasks. It’s shaping your thinking. It functions less like a digital intern and more like a second operator sitting beside you, pushing back.
Five Ways to Make AI Prompt You (Not the Other Way Around)
1. Project Partner
Most people: “Draft a project plan for this initiative.”
Better approach: “Here’s my draft project plan. Ask me 15 uncomfortable questions as if you were a skeptical executive, a risk officer, and an annoyed customer. Don’t suggest solutions yet, just challenge me.”
You answer those questions conversationally. Then you ask the LLM to refine your plan.
The result: You uncover failure modes before they become expensive. Your final plan is battle-tested, not just nicely formatted.
2. Meeting Prep Interrogator
Most people: “Create an agenda for my partner call.”
Better approach: “I have a 30-minute call with [stakeholder]. My rough goal is X. Ask me questions to clarify: What success really looks like. What I cannot afford to happen. What trade-offs am I willing to make? Keep asking until you can propose a precise agenda.”
By the time the LLM generates the agenda, you’ve already clarified your priorities and negotiation posture. You go in less scripted and more prepared.
3. Execution Guardrail for Repeatable Workflows
Most teams have recurring workflows: client onboarding, campaign launches, and audit prep.
Most people: “Create a checklist for onboarding a client.”
Better approach: First, generate the checklist. Then: “For each step, prompt me with: ‘What has changed since last time?’ ‘Where could this fail?’ ‘Who else should be consulted?'”
You’ve turned a static checklist into an interactive guardrail. The LLM repeatedly asks you to adapt and think, reducing copy-paste mistakes and context-blind decisions.
4. Strategy Challenger, Not Strategy Writer
Most people: “Write a GTM strategy for this product.”
Better approach:
- “Act as a skeptical investor. Ask me 10 questions that would make you decide not to fund this.”
- “Now act as a regulator. Ask questions that would make you worried about compliance.”
- “Now act as an early adopter. Ask what I’d need to believe to switch to this solution.”
You respond to each wave. Then ask: “Using my answers, synthesize a GTM strategy that directly addresses these concerns.”
The final output isn’t a generic template. It’s a document that has survived rigorous interrogation from multiple angles.
5. Deep Learning Accelerator
When entering a new market or regulatory space, most people ask: “Explain X to me” or “Summarize these guidelines.”
Better approach: “Quiz me on the basics of X. Track where I hesitate or answer incorrectly. Then ask application-level questions. Then ask me to critique my own answers.”
You’ve used AI not just to transfer information, but to build layered understanding: recall → application → self-critique. That’s how knowledge becomes usable judgment, not just notes in Notion.
The Simple Habit Shift That Changes Everything
You don’t need complex agents or custom platforms to start.
You need one habit:
Before asking your LLM for an answer, ask it to ask you better questions.
Some patterns I use:
“Interrogate before you generate” “Before drafting anything, ask me 10 clarifying questions about my goal, audience, constraints, and success metrics.”
“Play roles I can’t easily access” “Prompt me like a regulator/investor/angry customer would. Stay in character.”
“Push me up the thinking ladder” “Don’t let me stay surface-level. Keep asking ‘why’, ‘what if’, and ‘what would change if you were wrong?'”
“Help me institutionalize good judgment” Turn your best questions into reusable templates for your team: onboarding scripts, pre-mortem scripts, risk reviews. Your organization builds a question bank as an asset.
The Real AI Advantage Isn’t Speed. It’s Judgment.
In a few years, everyone will have access to similar models.
The advantage won’t be “who can get AI to write faster.”
It will be:
- Who consistently exposes their own blind spots before the market does
- Who makes higher-quality decisions under time pressure
- Who builds teams that think with discipline, not just act with speed
That advantage comes from treating LLMs not as vending machines for content, but as intelligent systems that challenge you before you hit send, launch, or commit.
So the next time you open an AI tool, resist the urge to fire off another “write this for me” prompt.
Instead, hand it the pen and say:
“Your job is not to answer me yet. Your job is to ask me the questions that will make my eventual answer impossible to ignore.”
That’s the quiet upgrade: from “getting more done with AI” to “becoming the kind of professional whose thinking is multiplied by it.”
And in the long run, that’s the only multiplier that really compounds.
Read my next article on why being polite to AI get’s you worse results.